I have a question as to how one should interpret clinical report records with “WITHDRAWAL” as the clinicalStage value. Intuitively, one might think that this reflects a no longer approved drug-indication association, but from what I can see, most of these entries reflect withdrawals of marketing authorisations (commercial withdrawal), hence not related to regulatory/safety aspects. Could you comment on the strategy for surfacing withdrawal data, and also whether the withdrawal applies to a specific product rather than the compound (e.g. “Docetaxel Teva” versus docetaxel).
that is a very good question, and something that I am afraid is not clearly stated in the documentation.
We think of clinical reports as evidence derived from clinical records, and as you point out, different records represent different types of clinical evidence at different levels of granularity. In practice, all withdrawal sourced from EMA Human Drugs come from evidence of medicinal products rather than compounds, so the annotation here is not on a per-drug basis. This means that querying clinical reports where clinicalStage == WITHDRAWAL won’t give you exactly what you’re looking for.
If you want to get a curated list of drugs that have been withdrawn across indications, check the Drug Warning dataset instead, filtering by warningType == Withdrawn to get a list of all withdrawn drugs according to ChEMBL. Note that these withdrawal records are also integrated into the clinical reports dataset, but are not straightforward to identify.
I hope this is helpful. I’ll try to make the documentation clearer on this matter.